Today's read
The standing market regime is choppy (issued 16 Oct, matures 23 Oct). We are not issuing a second regime row while that one is still running; a fresh regime read is due once it matures.
This run adds two new one-day calls. Both lean lower, and both rest mostly on price momentum:
- Ethereum (ETH): leans to reach −1.23% within the next day, at roughly 61%, against a ~53% historical touch base rate for a move that size. Coverage in the two-day window leans bullish on ETH, so this lean comes from momentum and runs against the coverage tone.
- Pump.fun (PUMP): leans to reach −2.96% within the next day, at roughly 66%, against a ~62% base rate. Coverage this window leans bullish on PUMP, including an analyst bull case, so the downside lean is again momentum-led and runs against the coverage. The gap over the base rate is small.
A touch call is graded on whether price reaches the stated level at any point inside the horizon. Where it closes does not count. The complement of a call is "does not reach the level", not "moves the other way".
Not issued, and why
- XRP, SOL: both cleared the conviction gate on every horizon, but each coin already carries two standing calls (the per-coin ≤2 limit).
- ETH 7d: the standing 7d call is materially unchanged, so nothing new was issued.
- PUMP 7d: the 7d book has 6 calls. The open slot is reserved for an idiosyncratic read, meaning one driven by the coin's own coverage. PUMP's lean comes from momentum while its coverage leans the other way, so it does not qualify.
- AVAX, DOGE, ADA, SUI, BNB, UNI, ENA, JUP: no conviction.
- HYPE: the engine could not give a probability (too little price history).
- Structural and relative questions: the new-ATH and altseason questions each already carry two standing reads. The drawdown scan gave no emit. The SOL and XRP relative reads re-scored within 0.10 of their standing numbers, so nothing new was issued.
Receipts: resolved since the last run
All three one-day calls issued on 21 Oct missed. Each one stated an upside level, and each coin fell instead:
| Call | Stated | Issued | Matured | Resolved | Outcome |
|---|---|---|---|---|---|
| NEAR: leans to reach +3.39% within the next day | 68% | 21 Oct 21:30Z | 22 Oct 21:30Z | 22 Oct 22:34Z | Miss: level not reached, closed −8.16% |
| Zcash: leans to reach +2.30% within the next day | 59% | 21 Oct 21:30Z | 22 Oct 21:30Z | 22 Oct 22:34Z | Miss: level not reached, closed −9.75% |
| Bitcoin: leans to reach +0.88% within the next day | 62.5% | 21 Oct 21:30Z | 22 Oct 21:30Z | 22 Oct 22:34Z | Miss: level not reached, closed −3.82% |
This run: 0 hits, 3 misses.
How the model is scoring (trailing 30 days to 22 Oct 2026, touch-graded calls only)
- 7-day calls: Brier Skill Score vs the base rate is −0.0034 (n=10). Our published probabilities have not yet beaten quoting the historical base rate. Reliability is 0.0375 (lower is better; this is where overconfidence shows) and resolution is 0.045 (higher is better).
- 1-day calls: Brier Skill Score is −0.1495 (n=29). That is clearly worse than the base rate, and today's three misses add to it.
- Directional-accuracy note, not a skill measure: on the 7-day pool, hit rate minus the mean base rate is −0.0113.
- The 7-day pool is below the 25-call minimum, so the verdict is insufficient data. These probabilities are uncalibrated and provisional.